Bifalgorithm post

There's so many machine learning algorithms. Here's how you can learn all of them. This is day four of trying to become cracked at machine learning in 30 days. Scikit-learn has an awesome cheat sheet that shows you all the important algorithms to know and use based on different tasks. Over the next few days, I'm going counterclockwise through each category, starting today with regression. On day two, I talked about basic linear regression. At its core, the model tries to minimize the average square difference between its predictions and the actual values. But with larger data sets, calculating every example at once can be slow, so instead we can update the model using one example at a time. That's called stochastic gradient descent. Another challenge is overfitting, which is when the model fits the trading data too closely and doesn't generalize well to new data. One way to prevent this is regularization, which adds a penalty for large weights. There's two methods called L1 and L2. Lasso regression uses L1 regularization, ridge regression uses L2, So regularization unlocks three more algorithms. Follow to stay on this journey and see you for day five.

There's so many machine learning algorithms. Here's how you can learn all of them. This is day four of trying to become cracked at machine learning in 30 days.…